Preserve '*' when formatting a 'SELECT *' statement

[agavra]

Description

Consider stream X with fields A and B. Previously, the following semantics would hold:

// Statement statement = resolve("SELECT * FROM X");
print(SqlFormatter.formatSql(statement));
> SELECT X.ROWTIME "ROWTIME", X.ROWKEY "ROWKEY", X.A "A", X.B "B" FROM X X;

The new behavior replaces all of the select items that came from a * with *, for example:

// Statement statement = resolve("SELECT *, A AS ONE FROM X");
print(SqlFormatter.formatSql(statement));
> SELECT *, X.A "ONE" FROM X X;

Motivation

In order to support #2436, we need to make sure that SELECT * statements do not expand to include ROWKEY/ROWTIME. Fixing the root cause of this behavior is tricky and perhaps backwards incompatible (see discussion #2436 (comment)), so in the meantime I propose the behavior in this PR, which I think is good independent of that PR.

Testing done

New unit tests.

Reviewer checklist

• Ensure docs are updated if necessary. (eg. if a user visible feature is being added or changed).
• Ensure relevant issues are linked (description should include text like "Fixes #")

[rodesai]

LGTM. One semi-nit inline. To clarify, this doesn't actually fix the semantics of '*' - its just deferring the final evaluation to the executor.

[rodesai]

Can we rename this from getSource to something else (maybe getAllColumns) ? It's confusing because source generally refers to a data source (stream/table) in the parser/engine. Same for other usage of the word source.

[spena]

Agree with @rodesai. source refers to the '*' column only, so it makes more sense to rename this method (perhaps the variable name as well) to one that relates to it, like starColumn, allColumns, whatever.

[spena]

The code looks good. Just the minor feedback about the StatementRewriter.getSource() method name.

[spena]

Agree with @rodesai. source refers to the '*' column only, so it makes more sense to rename this method (perhaps the variable name as well) to one that relates to it, like starColumn, allColumns, whatever.

[agavra]

Thanks for reviews @rodesai @spena. About your clarification - you are correct, this does not fix the semantics (that's a much more involved change).